The job search, in data · as of 2026-09-16
49 applications since April. 4 human conversations. 1 portfolio review. 0 offers. Companies are anonymised; the numbers are not. The fun is in the craft; the point is transparency. A point-in-time diagnostic of a live search, not a verdict on it.
01 · Where applications die
27 applications reached a CV screen. 4 became a conversation with a human. The screen is where this search dies, before anyone has seen the work.
02 · How fast they die
A same-day rejection is a filter, not a decision. Filled dots had a human involved at some point; hollow ones never did. 11 applications were never answered at all.
A human was involvedNever any humanStill open
* month-precision records: the log dates most early entries by month only.
03 · Does my own fit assessment predict anything?
Almost flat. A 5/5 fit (#9) and a 4.5/5 (#40) died at the CV screen like the 2/5s. The one application that reached a portfolio review (#33) was a 3.5. Whatever the screen measures, it is not fit.
Closed (rejected or ghosted)Still open
* the fit score is my own assessment, scored when I applied, not an objective measure. The method note at the foot of the page says how it is built.
04 · Channel against furthest stage
42 cold applications produced 2 conversations. The process that went furthest began with the company reaching out in September, five months after it had rejected the same profile’s cold application in April. The sample sizes make this a hint, not a statistic.
05 · Self-audit: blockers I knew about and applied anyway
36 of 49 applications carried at least one blocker I knew about at submission. This chart judges me, not the market, and it is the only one here that could change behaviour.
06 · Did better materials change anything?
The CV was rebuilt twice and a new portfolio shipped, and the line does climb. But split it by channel and the climb disappears: since August every human response has come through an intermediary or an inbound recruiter. The cold channel’s September rate is zero. Better materials did not fix the channel; a different channel did.
07 · Geography and outcome
Solid is rejected, muted is ghosted, teal is still open. No geography behaves meaningfully better: the screen is the screen everywhere.
RejectedGhostedStill open
08 · Ghosting over time
11 of 49 applications, about 22%, were never answered. Recent months read low only because ghosting takes two months to earn its name.
Share never answeredToo recent to count
* too recent to have earned the label.
09 · Domain sprawl
49 applications went to 30 different domains, and 20 of those domains got exactly one. The most common known blocker was domain mismatch, and the only specific rejection feedback named domain fit. The sprawl and the rejections are the same fact, seen from two sides. This chart argues for concentration.
Read the other way, it is a rough map of who was hiring senior designers in the AI era of 2026: health, money, security, housing, and tools for developers. Filtered through my own choices, so a sample of one search, not a census of the market.
1 application each
AI talent · Community SaaS · Construction · Consumer apps · Consumer health · Creative tools · Design tools · E-commerce · Edtech · Employee benefits · Energy · Gambling · Home services · Instruments · Insurance · Marketplaces · Martech · Mobility · Research tools · SaaS
10 · Did I get pickier?
Yes. The average fit score of what I applied to climbed every month. The selection tightened; the cold channel’s response did not follow. Discipline improved on my side of the screen only.
* same fit score as chart 03: mine, subjective, scored at submission.
Method
Every application is logged with dates, stage reached, channel, my own fit score, and blockers known at submission. Companies and people are removed at extraction, before this page is built, and a test fails the build if a name reaches it. Three roles assessed but never applied to are excluded from every rate.
The fit score deserves its own disclaimer: it is subjective by design, my own reading of the match, not a property of the role. It is built the same way every time. Before sending, each application gets a written assessment, strengths, gaps, and why I applied, and that assessment is condensed into a score from 1, poor, to 5, exceptional, in half steps. It is scored with what I knew at submission, and the outcome does not revise it afterwards.